Detecting malignancy in PSC

SOC in other patient categories, visual assessment as been reported to be the most sensitive method to detect CCA (sensitivity 95%, specificity 75%) (206). In our series we used the first generation fiber-optic Spyglass system with image quality often suboptimal and easily

impaired by mucus, blood or PSC alterations in the bile duct. Since the study period, a newer system has been developed with a digital image solution (SpyGlass DS; Boston Scientific) and higher image quality. However, even with instruments with superior image quality to the first generation Spyglass system the detection of dysplasia seems difficult in PSC. In a case-series of 30 PSC patients, high-resolution per-oral video cholangioscopy with narrow-band imaging (NBI) was evaluated for detection of CCA and dysplasia (231). With this method, CCA and tumor margins could be detected, but visual assessment of the bile duct mucosa did not increase the detection of dysplasia.

Three patients were diagnosed with CCA during follow-up. The first patient was diagnosed at inclusion and had endoscopic appearance, biopsies and brush cytology results positive for malignancy. The two remaining patients were diagnosed after 17 months with and

intrahepatic CCA and after 35 months with a hilar CCA. We included these patients when evaluating diagnostic accuracy for SOC-guided sampling, although this approach can be criticized. Sensitivity, specificity, PPV, NPV and accuracy were 33% (95% CI 1-91%), 100%

(95% CI 92-100%), 100% (95% CI, 3-100%), 95% (95% CI 99%) and 96% (95% CI 85-100%). The few number of events however precludes any firm conclusions based on these findings.

5.2.1.3 Adverse events

The frequency of previously reported ERCP-related adverse events in PSC vary considerably (232). We found a relatively high risk of procedure-related adverse events in this patient series. In 7/47 patients (14%) complications occurred. In six cases, the adverse events were classified as mild with little impact or prolonged hospital admission no longer than 3 days.

One case of cholangitis was classified as moderate. Although the few number of patients limits conclusions we argue that risks should be carefully weighed when evaluating patients with this invasive technique.

5.2.1.4 Additional technical value

We also tried to evaluate possible technical advantages associated with SOC. First, one possible advantage is guidance through visualization of the bile duct to pass the guide-wire through narrow strictures. This was found to be the case in 4/45 (9%) patients. Second, it has previously been described that SOC increases the detection and subsequent removal of bile duct stones in PSC (229). In 12/45 (27%) patients in our series biliary stones were detected and removed in 11 of these cases. The clinical implication of these possible advantages remains to be determined. To access and sample suspicious lesions in PSC patients can be technically demanding. SOC could, in this setting, offer a technique to reach and sample otherwise inaccessible strictures when necessary. Whether increased detection and removal

of biliary stones in PSC is beneficial or not is not known. These hypothetical advantages deserve attention in future studies on the subject.

5.2.1.5 Strengths and Limitations

The main strength of this study is the prospective design that allows for a robust evaluation of success rates and adverse events. However, given the small sample size results should be interpreted with caution. Also the limited number of patients with malignancy precludes any meaningful evaluation of the diagnostic accuracy of SOC with targeted biopsies in this patient-series.

5.2.2 Biliary brush cytology and FISH

In this retrospective study (Paper III), we evaluated the diagnostic performance of brush cytology with sequential use of FISH in equivocal cases for detection of CCA and biliary dysplasia in patients with PSC. We showed that a correct diagnosis could be made in 95%

(197/208) of cases in this relatively unselected cohort of PSC patients with a clinical indication for ERCP. We also demonstrated that sequential use of FISH in equivocal cases shows a higher sensitivity (80%), than previously reported results of routine cytology, while an acceptable level of specificity (96%) was maintained (192). Furthermore, we found that FISH was highly sensitive (100%) for the detection of CCA in PSC patients with no evidence of malignancy on prior imaging and that false positive results correlated to biliary dysplasia in explanted livers.

5.2.2.1 The positive predictive value of FISH

We believe that the present study, complements the current data on the value of FISH analysis in PSC. Previous studies in the field have focused on evaluation of the risk for development of CCA in patients with a positive FISH analysis using a longer follow-up time.

In one study of 102 PSC patients with no definite signs of CCA on imaging and equivocal cytology results, 29% was diagnosed with CCA within 2 years. The risk of CCA was increased more than eight times in cases with a positive FISH (HR 8.70 (95% CI 3.79–

19.99). In another study of 235 PSC patients, in which patients were followed up to 34 months, the sensitivity of FISH for detection of later CCA development was shown to be 46% and the specificity 88%. At Karolinska University Hospital, as well as in all Nordic liver transplant centers, biliary dysplasia is an indication for LTx in PSC. This transplant setting, in which patients with HGD are offered LTx, has allowed us to evaluate the diagnostic accuracy of biliary brush cytology with FISH in a relatively short perspective (within 12 months). We therefore chose a diagnostic study design in contrast with multiple previous studies

evaluating FISH as a risk factor for developing CCA at long-term follow-up (196, 199, 201, 202, 233). The diagnostic perspective in the present study gives further guidance how to interpret brush cytology in PSC patients.

As anticipated, applying FISH in equivocal cases of routine brush cytology showed a higher sensitivity than previously published results of routine cytology (192), but at the cost of

decreased specificity for identifying established CCA. The PPV of a brush positive for either malignant cells or FISH was 60% in our study. This may in part be explained by previously published results showing that the genetic alterations detected by FISH are not specific for CCA but also present in preneoplastic changes. In a study of by DeHaan et al., biliary dysplasia in liver resections and explanted livers from PSC patients, demonstrated either FISH polysomy or homozygous 9p21 loss in all specimens (5/5) (234). Thus, it is likely that, at least partly, the positive findings of FISH in this study are explained by biliary dysplasia.

When including HGD and both HGD and LGD as positive endpoints, PPV increased from 60% for CCA only to 75% and 90%. Although the number of patients with dysplasia was low in the present study, all patients with HGD in explanted livers (3/3) had a brush cytology positive for FISH. Previously published data strongly supports that CCA develops in areas with dysplastic lesions in PSC, but the time and course of this process is not known (186). In addition, PSC patients positive for FISH polysomy, either in equivocal cytology or in

multifocal or serial samples, are highly likely to develop CCA (196, 199, 201, 202).

Chromosomal aberrations detected by FISH may itself indicate a higher risk for malignant transformation than histologically identified dysplasia. Although probable, it is, however, difficult to obtain figures supporting this assumption. Risk stratification of PSC patients according to FISH status might be a rational approach although we note that cancer was already prevalent in 58% of patients with equivocal brush cytology with positive FISH and a negative imaging (i.e. no definite malignancy).

5.2.2.2 The negative predictive value of FISH

A high NPV of 98% was reached using FISH in equivocal cases. In clinical practise, it is important to rule out the presence of CCA or HGD in a significant stricture seen at imaging.

Our study population includes referral cases from other regions but most patients were included from our primary catchment area. This is reflected by the relatively low prevalence of CCA of 7%, and that only two patients had obvious signs of malignancy prior to brushing.

Thus, a high NPV for the entire group of patients with clinical indication for ERCP is expected. However, even in patients with equivocal cytology a negative FISH result was highly accurate for ruling out malignancy at 12 months (NPV 100%).

5.2.2.3 Strengths and limitations

Strengths of this study include the relatively large size of the study cohort covering all PSC patients who underwent biliary brushings during the study period. Also, almost all identified patients had a complete follow-up and could be included in the final analysis thus limiting selection bias. There are, however, several limitations to our study, mainly due to the

retrospective design with its inherent flaws. During the study period, FISH results have been used in clinical decision-making, introducing verification bias, meaning that patients were not randomly allocated to interventions (e.g. LTx, surgical resection), which are of importance for the final diagnosis. The consequence of this partial verification would be a potential overestimation of the prevalence of early CCA and biliary dysplasia in patients with a

positive FISH. Consequently, an overestimation of the NPV would then likewise be the case since patients with a negative FISH finding might harbor undetected early CCA or HGD.

Furthermore, since the results of biliary brush cytology are available to the pathologist interpreting the final histopathology (e.g. from an explanted liver) diagnostic review bias might also have affected the results in a similar fashion.

The choice of using only the index brush cytology result for calculation of the diagnostic accuracy is also debatable. The main limitation of this approach is that it also includes false negative results by sampling error, which is separate from the cytological diagnostic method itself. However, when we included results from repeated brushings the values for diagnostic accuracy did not change, although the number of cytology results with insufficient material did decrease (data not shown).

The follow-up time was restricted to 12 months. We argue that this approach is relevant for evaluation of FISH as a diagnostic test since a reduced follow-up time reflects the prevalence of CCA or dysplasia at the time of first brushing rather than progression of preneoplastic changes that might occur over time. Finally, there were a limited number of patients with CCA in our cohort, despite having a relative large overall sample size. Our results should therefore be interpreted with caution.

5.2.3 Present and future perspectives

Detection of malignant and premalignant lesions in the bile ducts of PSC patients remains a challenge. The use of FISH and SOC both provide additional diagnostic means that together and in combination with other techniques such as imaging can improve diagnostics. Studies with the aim to investigate biomarkers in serum and bile are ongoing and would further have the potential to improve diagnostics. However, the rarity of PSC makes it difficult to evaluate such markers and robust diagnostic methods cannot be expected in the near future.